Recently, I performed a study in which I needed to test for dual mediation – when the
indirect effect of Y on X goes through two mediators
(pictured here). Whenever I need to test for mediation or moderation of any type, I first go to Hayes’s PROCESS macro for SPSS. It is easy, and it adheres to most all
current best-practices. If you don’t have it yet, I STRONGLY suggest that you download it now: http://www.processmacro.org/index.html . It has saved me hours (days?) on statistical analyses.
So, I opened up the documentation for PROCESS and I found…that it does not have a model for dual mediation! Of all things, I would have assumed that PROCESS would be able to test for dual mediation!
I searched online for solutions, and I even referred to mediation textbooks to
devise a formula to test for dual mediation. After having moderate success, I
finally went back to PROCESS. Out of frustration, I set the program to test the traditional mediation model (with one mediator), but I included two mediators in the dialogue window. And what did I see? It provided output for dual mediation!
While this information may not be groundbreaking, it certainly helped me. I hope that it can also help someone who is as frustrated with dual mediation as I was. The permanent page for dual mediation testing can be found here: Testing for Dual Mediation .
If you need any help running dual mediation in process (or any other statistical questions), feel free to email me at MHoward@SouthAlabama.edu. I always
enjoy hearing from people running interesting statistics!